A New Model

Where I work has been going through a restructure and as of next year we will have a new operational model. As yet there has been no real details of what that may look like, however it has made me think about the changing landscape of how the business of analytics is conducted. Throughout my career I’ve seen the same challenges persist without any real effort to have them addressed, the main reason for this in my opinion is a gap between decision makers and the people on data front-line. At the start of my career this didn’t seem to be an issue, but this was before the data revolution had really begun. Now though everyone wants to get on board the data driven train, however making this happen effectively and efficiently is something that not many organisations have mastered, but why is that? Let’s start by asking one question – How many of you have a CAO, CDO or a CDO in your organisation? If you don’t have any of these chances are your organisation isn’t taking analytics seriously. In a parallel universe where I was in charge of putting an analytics dream team together, below is what I would choose.

Step 1: Get a Data Developer

Data Scientist are all the rage at the moment, but in my experience unless organisation is very mature I would say it is unlikely the top priorities are predictive analytics, maybe you enhance forecasting models, but in my opinion more value could come from a Data Developer. A Data Developer is someone who like an ETL developer but is very business focused and envision how to unlock and build new datasets that the data analysts can use.

You may think oh yes we have a person in IT like that. Sure but how long does it take to have requests serviced by IT? Too long. Get this person in your analytics team and have them create prototypes and pilots, then push the beta versions of the new data assets out to IT to have them properly integrated into the data warehouse.

This role also provides the translation of the workings of data systems to business processes.

https://hackernoon.com/data-developer-d20132529c80

Step 2: Get a Visualization Specialist

Ok so this is an area that we all think we know about, but I have met very few people with an analytics background that know any theory about what is arguably the most important part of analytics, that being the delivery of information.

Ok so the preferred skill set for this role is pretty board and the preferred skill set seems to be split analytics visualisation role and UI/front end roles. What you want ideally though is someone that has some level of (Graphic) Designer experience. People with design experience know how your page should be laid out, the importance of colour, can whip up an eye catching infographic etc. If you’re really lucky they will also be able to distill complex datasets into awesome interactive web charts using for example can create interactive visualisations built on D3.js.

https://medium.com/@lynn_72328/data-visualization-versus-ui-and-data-science-d59182d58af4

Step 3: Get a Test Driven Analysis Lead

Test Driven Development is fast becoming accepted as a best practice approach for software engineering. And although TDA (or TDDA) is probably just a concept that is seldom used in industry at the moment, my prediction is that as analytics further matures some form of TDA will be adopted.

Step 4: Get a Cloud Architect

After spending two years managing an in house analytics environment I now have an understanding and appreciation of what it takes to operationally manage and maintain an analytics environment. I could write a whole separate post on the challenges I faced, but the takeaway is that with cloud platforms you can offload responsibility and gain so much more flexibility.  

Step 5: Get a Data Champion 

Yes I agree it is a bit of a cringe worthy term, but effectively it means that you have someone high up in the organisation that understands what data you have and can spread the good word and can explain when opportunities arise how to use the data effectively to solve business problems.

https://blog.agilenceinc.com/genesis-of-the-data-champion

Step 6: Get a Delivery Manager

This is role is really important. Requests for data projects never end so it really easy for teams to get burnt out. The Delivery Manager should essentially be that force field around your team as the go between to the outside world and managing external stakeholder expectations. On top of that the role should remove roadblocks, for example gaining access to subject matter experts.

Step 7: Data Governance – a team effort (enforced by the C level)

Initially I thought a Data Governance Manager should be part of the team, but then soon after beginning to research I stumbled on an article with this comment:

Data governance focuses on the aftermath of poorly designed data, ill-conceived business processes which results in a reactive form of data management.

What most organizations are lacking is Data Literacy. It’s like trying to govern books when no one knows how to read…

…The first step in any data governance initiative should be to learn about metadata. Create a comprehensive business glossary and data dictionary for the data. This best practice alone will result in a deep understanding of the metadata and will expose many of the most challenging issues an organization will face when attempting to govern data. What will be discovered are ambiguous definitions, lack of naming standards, a lack of semantic understanding of the data and discovery of the various interpretations of metadata.

That comment (thanks Richord1) embodies everything I have ever witnessed. It further made me think that governance shouldn’t be the responsibility of a single role, instead all team members should be invested in updating and maintaining metadata catalogues. Of course metadata seems to be always lumped into the documentation category – doesn’t exist because people hate doing it. So for effective governance this must be understood and made a priority by one of your executives.

Final thoughts:

Well that’s my take on the roles necessary to run a modern efficient data analytics team, what stood out to me at the end of writing this post is that I assembled a team that looks a lot like an agile software development team. The only obvious drawback to me is that your non-technical analysts may feel a little intimidated. Both technical and non technical members need to embrace each others different skill sets and view of the world. The best teams are those that can bridge this gap, as the sum of the whole is more than its individual parts. Leave your ego at the door and be kind to one and another.

https://www.entrepreneur.com/article/249781